def test_long_to_wide_conversion(self, longdata):
g = TimeVaryGFormula(longdata,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
lf = g._long_to_wide(longdata, id='id', t='t')
expected_lf = pd.DataFrame.from_records([{
'A_0': 0,
'A_1': 1,
'A_2': 1,
'Y_0': 0,
'Y_1': 0,
'Y_2': 1,
'W_0': 5,
'W_1': 5,
'W_2': 5,
'L_0': 25,
'L_1': 20,
'L_2': 31,
'id': 1
}]).set_index('id')
pdt.assert_frame_equal(lf,
expected_lf[[
'A_0', 'A_1', 'A_2', 'Y_0', 'Y_1', 'Y_2',
'W_0', 'W_1', 'W_2', 'L_0', 'L_1', 'L_2'
]],
check_names=False)
def test_error_covariate_label(self, sim_t_fixed_data):
g = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
time_in='t0')
with pytest.raises(ValueError):
g.add_covariate_model(label='first', covariate='W1', model='W2')
def test_error_monte_carlo1(self, sim_t_fixed_data):
with pytest.raises(ValueError):
TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='W1',
time_out='t')
def test_error_estimation_method(self, sim_t_fixed_data):
with pytest.raises(ValueError):
TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='W1',
time_out='t',
method='Fail')
def test_error_continuous_outcome(self, sim_t_fixed_data):
with pytest.raises(ValueError):
TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='W1',
time_out='t',
time_in='t0')
def test_error_sr_other_models(self, sim_t_fixed_data):
g = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
g.outcome_model('A + W1_sq + W2 + W3', print_results=False)
g.exposure_model('W1_sq', print_results=False)
with pytest.raises(ValueError):
g.fit(treatment='all')
def test_sr_custom_time_point(self, data):
g = TimeVaryGFormula(data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
g.outcome_model('A + L', print_results=False)
# values come from R's ltmle package
g.fit(treatment="all", t_max=2)
npt.assert_allclose(g.predicted_outcomes, 0.33492, rtol=1e-5)
g.fit(treatment="none", t_max=2)
npt.assert_allclose(g.predicted_outcomes, 0.51228, rtol=1e-5)
def test_sr_gap_time(self, data):
g = TimeVaryGFormula(data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t2',
method='SequentialRegression')
out_m = 'A + L'
g.outcome_model(out_m, print_results=False)
g.fit(treatment="all")
npt.assert_allclose(g.predicted_outcomes, 0.4051569, rtol=1e-5)
g.fit(treatment="none")
npt.assert_allclose(g.predicted_outcomes, 0.661226, rtol=1e-5)
def test_error_treatment_type(self, sim_t_fixed_data):
g = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
time_in='t0')
with pytest.raises(ValueError):
g.fit(treatment=1)
g = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
with pytest.raises(ValueError):
g.fit(treatment=1)
def test_sr_warning_outside_time_point(self, data):
g = TimeVaryGFormula(data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t2',
method='SequentialRegression')
g.outcome_model('A + L', print_results=False)
with pytest.warns(UserWarning):
g.fit(treatment="all", t_max=6)
npt.assert_allclose(g.predicted_outcomes, 0.33492, rtol=1e-5)
def test_sr_custom_treatment(self, data):
g = TimeVaryGFormula(data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t2',
method='SequentialRegression')
out_m = 'A + L'
g.outcome_model(out_m, print_results=False)
g.fit(treatment="g['t'] != 2")
npt.assert_allclose(g.predicted_outcomes, 0.48543, rtol=1e-5)
def test_monte_carlo_for_single_t(self, sim_t_fixed_data):
# Estimating monte carlo for single t
gt = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
time_in='t0')
gt.outcome_model('A + W1_sq + W2 + W3', print_results=False)
gt.exposure_model('W1_sq', print_results=False)
gt.fit(treatment="all", sample=1000000
) # Keep this a high number to reduce simulation errors
# Estimating with TimeFixedGFormula
gf = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
gf.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
gf.fit(treatment='all')
# Expected behavior; same results between the estimation methods
npt.assert_allclose(gf.marginal_outcome,
np.mean(gt.predicted_outcomes['Y']),
rtol=1e-3)
def test_error_sr_recurrent_outcomes(self):
df = pd.DataFrame()
df['id'] = [1, 1, 1]
df['Y'] = [0, 1, 1]
df['A'] = [0, 1, 1]
df['t'] = [0, 1, 2]
g = TimeVaryGFormula(df,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
g.outcome_model('A', print_results=False)
with pytest.raises(ValueError):
g.fit(treatment='all')
def test_sequential_regression_for_single_t(self, sim_t_fixed_data):
# Estimating sequential regression for single t
gt = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
gt.outcome_model('A + W1_sq + W2 + W3', print_results=False)
gt.fit(treatment="all")
# Estimating with TimeFixedGFormula
gf = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
gf.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
gf.fit(treatment='all')
# Expected behavior; same results between the estimation methods
npt.assert_allclose(gf.marginal_outcome, gt.predicted_outcomes)
def mc_gformula_check():
df = load_sample_data(timevary=True)
df['lag_art'] = df['art'].shift(1)
df['lag_art'] = np.where(df.groupby('id').cumcount() == 0, 0, df['lag_art'])
df['lag_cd4'] = df['cd4'].shift(1)
df['lag_cd4'] = np.where(df.groupby('id').cumcount() == 0, df['cd40'], df['lag_cd4'])
df['lag_dvl'] = df['dvl'].shift(1)
df['lag_dvl'] = np.where(df.groupby('id').cumcount() == 0, df['dvl0'], df['lag_dvl'])
df[['age_rs0', 'age_rs1', 'age_rs2']] = spline(df, 'age0', n_knots=4, term=2, restricted=True) # age spline
df['cd40_sq'] = df['cd40'] ** 2 # cd4 baseline cubic
df['cd40_cu'] = df['cd40'] ** 3
df['cd4_sq'] = df['cd4'] ** 2 # cd4 current cubic
df['cd4_cu'] = df['cd4'] ** 3
df['enter_sq'] = df['enter'] ** 2 # entry time cubic
df['enter_cu'] = df['enter'] ** 3
g = TimeVaryGFormula(df, idvar='id', exposure='art', outcome='dead', time_in='enter', time_out='out')
exp_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + cd4 + cd4_sq +
cd4_cu + dvl + enter + enter_sq + enter_cu'''
g.exposure_model(exp_m, restriction="g['lag_art']==0")
out_m = '''art + male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + cd4 +
cd4_sq + cd4_cu + dvl + enter + enter_sq + enter_cu'''
g.outcome_model(out_m, restriction="g['drop']==0")
dvl_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + lag_cd4 +
lag_dvl + lag_art + enter + enter_sq + enter_cu'''
g.add_covariate_model(label=1, covariate='dvl', model=dvl_m, var_type='binary')
cd4_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + lag_cd4 +
lag_dvl + lag_art + enter + enter_sq + enter_cu'''
cd4_recode_scheme = ("g['cd4'] = np.maximum(g['cd4'],1);"
"g['cd4_sq'] = g['cd4']**2;"
"g['cd4_cu'] = g['cd4']**3")
g.add_covariate_model(label=2, covariate='cd4', model=cd4_m,recode=cd4_recode_scheme, var_type='continuous')
g.fit(treatment="((g['art']==1) | (g['lag_art']==1))",
lags={'art': 'lag_art',
'cd4': 'lag_cd4',
'dvl': 'lag_dvl'},
sample=10000, t_max=None,
in_recode=("g['enter_sq'] = g['enter']**2;"
"g['enter_cu'] = g['enter']**3"))
gf = g.predicted_outcomes
gfs = gf.loc[gf.uid_g_zepid != gf.uid_g_zepid.shift(-1)].copy()
kmn = KaplanMeierFitter()
kmn.fit(durations=gfs['out'], event_observed=gfs['dead'])
kmo = KaplanMeierFitter()
kmo.fit(durations=df['out'], event_observed=df['dead'], entry=df['enter'])
plt.step(kmn.event_table.index, 1 - kmn.survival_function_, c='g', where='post', label='Natural')
plt.step(kmo.event_table.index, 1 - kmo.survival_function_, c='k', where='post', label='True')
plt.legend()
plt.show()
def test_complete_mc_procedure_completes(self):
df = load_sample_data(timevary=True)
df['lag_art'] = df['art'].shift(1)
df['lag_art'] = np.where(
df.groupby('id').cumcount() == 0, 0, df['lag_art'])
df['lag_cd4'] = df['cd4'].shift(1)
df['lag_cd4'] = np.where(
df.groupby('id').cumcount() == 0, df['cd40'], df['lag_cd4'])
df['lag_dvl'] = df['dvl'].shift(1)
df['lag_dvl'] = np.where(
df.groupby('id').cumcount() == 0, df['dvl0'], df['lag_dvl'])
df[['age_rs0', 'age_rs1',
'age_rs2']] = spline(df,
'age0',
n_knots=4,
term=2,
restricted=True) # age spline
df['cd40_sq'] = df['cd40']**2
df['cd40_cu'] = df['cd40']**3
df['cd4_sq'] = df['cd4']**2
df['cd4_cu'] = df['cd4']**3
df['enter_sq'] = df['enter']**2
df['enter_cu'] = df['enter']**3
g = TimeVaryGFormula(df,
idvar='id',
exposure='art',
outcome='dead',
time_in='enter',
time_out='out')
exp_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + cd4 + cd4_sq +
cd4_cu + dvl + enter + enter_sq + enter_cu'''
g.exposure_model(exp_m, restriction="g['lag_art']==0")
out_m = '''art + male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + cd4 +
cd4_sq + cd4_cu + dvl + enter + enter_sq + enter_cu'''
g.outcome_model(out_m, restriction="g['drop']==0")
dvl_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + lag_cd4 +
lag_dvl + lag_art + enter + enter_sq + enter_cu'''
g.add_covariate_model(label=1,
covariate='dvl',
model=dvl_m,
var_type='binary')
cd4_m = '''male + age0 + age_rs0 + age_rs1 + age_rs2 + cd40 + cd40_sq + cd40_cu + dvl0 + lag_cd4 +
lag_dvl + lag_art + enter + enter_sq + enter_cu'''
cd4_recode_scheme = ("g['cd4'] = np.maximum(g['cd4'],1);"
"g['cd4_sq'] = g['cd4']**2;"
"g['cd4_cu'] = g['cd4']**3")
g.add_covariate_model(label=2,
covariate='cd4',
model=cd4_m,
recode=cd4_recode_scheme,
var_type='continuous')
g.fit(treatment="((g['art']==1) | (g['lag_art']==1))",
lags={
'art': 'lag_art',
'cd4': 'lag_cd4',
'dvl': 'lag_dvl'
},
sample=10000,
t_max=None,
in_recode=("g['enter_sq'] = g['enter']**2;"
"g['enter_cu'] = g['enter']**3"))
assert isinstance(g.predicted_outcomes, type(pd.DataFrame()))